mimo indoor
TRANSCRIPT
-
7/29/2019 mimo indoor
1/12
Performance and Capacity analysis of MIMO system at 5 GHz and
60GHz in Indoor Environment
KIRTHIGA.S AND JAYAKUMAR.M
Communication Research group, Department of Electronics and Communication EngineeringAmrita Vishwa Vidyapeetham
Coimbatore 641112
INDIA
[email protected], www.amrita.edu
Abstract:- The Millimeter Wave (MMW) band much useful for wireless indoor communications as it offers alarge amount of license free spectrum. Blockage by walls and furniture limits the range indoors, the link rangefor indoors being 10m and outdoors 100m in the case of spatial reuse. The idea of dual-band integration of5GHz/60GHz conceived by IST-Broadway is considered. The physical link layer throughput achievable in
5GHz would be limited by the available bandwidth while the huge bandwidth available for 60GHz would makeit feasible for multi-gigabit link. Spatial multiplexing can offer large capacity gains if the spatial correlation islow. In this work, MIMO channel for 5 GHz and 60 GHz is modeled and the channel capacity is determined.The Triple Saleh Valenzuela model (desktop environment) is chosen as suitable channel model for MillimeterWave while IEEE 802.11n channel model B (small offices) is chosen for 5GHz. The power delay profileobtained on simulation gives the Rician factor and the RMS delay spread indicating multipath fading and time-dispersive channel. The performance is analysed with respect to bit error rate (BER) for various antennaconfigurations with transceiver distance of 3m. Consideration of Line of-Sight (LOS) component showsreduced BER in lower Eb/No range of 1 to 4 dB for 60 GHz compared to 5 GHz radio. This result makesMMW suitable for integration with 5 GHz, whose link is weak for short range communication. The bit errorrate (BER) is compared for 2x2, 4x4, and 8x8 for different equalization techniques namely Zero Forcing (ZF),Maximum Likelihood (ML) and Minimum Mean Square Error (MMSE) for the two cases, which in turn had
reinforced the fact of better performance of MMW. ML detection offers optimal error performance for 60GHz. The channel capacity is found to be 2 Gbps for 60GHz and 600 Mbps for 5GHz. The link budget forMMW is also analyzed.
Key-Words:- Millimeter Wave, Time of Arrival (ToA), Angle of Arrival (AoA), Cluster arrival rate, Rayarrival rate, Spatial correlation.
1. IntroductionUncompressed multimedia data transfer betweenwireless devices is expected to grow exponentiallyrequires large bandwidth. This requirement wouldonly be satisfied by millimeter wave technologyespecially in the unlicensed spectrum centered at60GHz. For the efficient use of this spectrumnecessitous the detailed study on channelperformance and characteristics. The data rateshave been increasing constantly from 1 Mbps infirst generation to 600 Mbps with 802.11n MIMOproducts. IEEE 802.11n working group began aVery High Throughput Study Group (VHT SG) toinvestigate technologies giving multi-gigabit link
throughput [1][2]. The achievable throughput in 2.5GHz and 5 GHz bands would be limited by thebandwidth limitation in these two bands. The usage
models of WLAN cover environments like sync-and-go, downloading movies or pictures fromcamera needs increasingly higher throughput as the
quality and resolution increases. The growing useslike internet telephony, music, gaming, and in-
home video transmission have in turn increased thenumber of WLAN users. This has started to strain
the existing Wi-Fi networks which therefore, needsa frequency band that gives huge bandwidth andless interference [3][4].
The Millimeter Wave (MMW) being anattractive option for designing the multi gigabit
links as they offer 7 GHz bandwidth of spectrumspanning from 57 -64 GHz. The MMW has made
the possibility of usage of Wireless USB, WirelessGigabit Ethernet. The large available bandwidthand O2 absorption present at 60GHz and other
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 415 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
2/12
intrinsic atmospheric attenuation make the MMWuseful for high data rates and spatial reuse [5]. The
range is limited by the blockage of waves by wallsand furniture, the link range for indoors being 10mand outdoors 100m in the case of spatial reuse. Thisproduces fast and fading statistics in channel. At 60GHz, free space path loss is much higher. This
property has led to the solution of integrating theexisting 5 GHz network with the 60 GHz band.
The issues regarding the use of MMW initiate withthe physical layer [6][7]. Channel capacity forindoor LOS spatial multiplexing was maximized byconsidering the singular values of the channelmatrix [8][10]. Also degrees of freedom increased
with more number of high singular values, whichled to independent channels thus increasing thedata rate. Antenna array with uniform linear
configuration was accounted to deal with highpropagation loss inherent with MMW. The limit on
the array length without compromising channelcapacity was studied [9][10]. The channelconsidered in the above work was 2-ray and 6-raymodel accounting for the LOS and NLOS path[10].
The primary challenge is to cover the range forWLAN. The blockages occurring due to the
environment decrease the coverage range of theMMW. This can be reduced by high gain antennas,high transmit power and sensitive receiver [11]. As
the wavelength of the MMW is small, number ofantennas can be packed on a small platform. Thiswill increase the gain and hence there will be anincrease in the link budget characteristics. Careshould be taken to decrease the side lobes in theundesired directions [12]. Another key challenge inthe Medium Access Layer (MAC) lies in providingthe link robustness. As MMW is susceptible to highattenuations through obstructions, link is easilybroken. In addition, the use of directional antennasmakes the link sensitive to slight movement of theobjects. To combat this, multipath MAC and
efficient contention based schemes needs to bedesigned for the MMW propagation [13] [14].The feasibility of the MMW is much higher than 5GHz band for short-range communication. The datarate of more than 1 Gbps can be achieved easily inMMW band with lower order modulation schemeslike BPSK without the use of many antennas,whereas in 5GHz , the transmit and receive
antennas has to be increased and higher modulationschemes like QAM has to be used [15] . Variousparameters like the link budget, channel capacityand its comparison with the Shannon limit wasperformed varying the communication range from5 -20 m for 60 GHz system [16]. The EIRP
achieved by the MMW is nearly 4dB higher thanthat achieved in the 5 GHz band. This is possible
for MMW because a number of antennas can beintegrated which increases the gain of the link.
Modeling indoor propagation environment iscomplicated by large variability in building layoutand construction materials. Another important
element of indoor wireless operation that should betaken into account is interference. Indoor path loss
can change dramatically with either time orposition, because of multipath present [17]. Thewideband of waves used in indoor applicationsincrease the sensitivity to delay spread. Site-specific and site-general modeling are the two
general types of propagation modeling. As the pathloss combines with other channel impairments likedelay spread, there arises a necessity to use
directional antennas to obtain reliablecommunications [18][19]. To obtain better signal to
noise ratio (SNR) in MMW band and in order toeffectively use frequency with space divisionmethod, the effect of antenna directivity has to beconsidered.
Performance analysis of 60 GHz using TripleSaleh Valenzuela (TSV) was carried in the earlierwork for MIMO system [20]. The analysis and
results of the above work are used in this paper.MIMO system performance for 60 GHz and 5 GHzis studied and analyzed [21]. The channel model
used for the MMW propagation is the TSV whichis a modified form of Saleh Valenzuela (SV)model. This is a merger of two-path model and SVmodel. This model was contributed by NICT, Japanto 802.15.3c [22]. The complex impulse responseobtained from this model contains the Time ofArrival information (ToA), Angle of Arrivalinformation (AoA), and amplitude of each ray inthe cluster[20]. For 5 GHz, IEEE 802.11n channelmodel B is used. The performance of the systemswith respect to BER in lower and higher SNRranges, the channel capacity and link budget for
MMW is analyzed. In this paper, section 2 and 3discusses the channel models used for 60 GHz and5 GHz, section 4 discusses MIMO system forindoor environments using TSV model, which is acluster based model newly attempted in this workas opposed to ray tracing technique along with thevarious equalization techniques and section 5discusses the results with respect to BER
performance, computation of channel capacity of60 GHz and 5 GHz and the link budget analysis forMMW system.
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 416 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
3/12
2. Triple Saleh Valenzuela Model for
60 GHzIndoor channel modeling is found to be difficult by
the large number of obstructions, variations inlayout of the building and movements of people.
To satisfy all these, site-general models arepreferred. As the frequency of operation formillimeter waves is very high, large bandwidths areavailable for the link but this high frequency alsoleads to high path loss [22][23].
The TSV is a site-general model that merges S-V and two-path models. The impulse responsegives the information of the relative power of thefirst ray that arrives to that of the last receivedsignal component. The amplitude factor of thisimpulse response in TSV model is determined bythe distance between the millimeter device
positions and the heights of the antennas. Theuncertainty and high vulnerability of the deviceposition and the fading caused is modeled in TSVby the random variables generated by Poisson andLaplace distributions [17].
The impulse response of the modified SV modelaccounted for the antenna effects such as anglespread in the transmitter and receiver antenna [14].MMW being high frequency waves, both LOS andNLOS components tend to dominate. Hence thecomplex impulse response (CIR) of the TSV modelis given [7] as
)()()()( ,0 0
,, mll
l m
mllml Tttth +=
=
=
(1)
where is the direct wave component that holdsthe information about the heights of the transmitterand receiver antenna, distance between the antenna,reflection co-efficient and the wavelength at the
center frequency[17], l,m is the complex amplitudeof each ray, t is the time, Tl is the delay time of the
l-th cluster l,m is the delay time of the m-th ray in l-th cluster, l,m is the angle of arrival of the l-thcluster, l,m is the angle of arrival ofm-th ray in the
l-th cluster.The direct path component is generated by using
a two-path model as shown in [7]
+=
D
hhjGGGG
D frtrt
D 2102211
22exp
(2)
where D is the distance between the transmit andreceive antennas and h1 , h2 are the antenna heights
f, o, and D are the wavelength at the centrefrequency, the reflection coefficient, and theaverage distance distribution respectively. Gt1 andGr1 are the transmitter and receiver gains for thedirect path, and Gt2 and Gr2 are the transmitter and
receiver gains for the reflected path. The value of is very sensitive to small antenna movements [22].
The arrival rate of clusters and rays are defined bythe Poisson process as following
P(Tl | Tl-1) = exp{- (Tl-Tl-1)} l> 0
(3)
P(l | l,(m-1)) = exp{- (l-l,(m-1))} m > 0
(4)
where and are the cluster and ray-arrival raterespectively.
3. Channel model for 5 GHzThe channel used is IEEE 802.11n B that uses
5GHz as the centre frequency and 48 MHzbandwidth. The IEEE 802.11n group has developed
channel models applicable for indoor WLAN
systems. These models are developed by the clustermodels given by the Saleh-Valenzuela channel
models. The number of clusters considered for themodels varies from 2 to 6. Angle of Arrival (AoA),
Angle of Departure (AoD), Angular Spread (AS)vary for each cluster models. Cluster angularspread was found to be between 20
0and 40
0range
and mean AoA had uniform distribution. Thechannel correlation matrix can be obtained from all
these values of AoD, AS, AoA. The size of thechannel matrix depends on the number oftransmitter and receiver antennas. If the number of
transmit antenna is nt and that of receiver antenna isnr. the size of the channel would be nt x nr[18][19].
The MIMO matrix formulation is done by theequation as follows:
++
+= RayleighH
KRicianH
K
KPH
1
1
1
(5)
=
++
+ 2221
1211
1
1
2221
1211
1 XX
XX
Kj
ej
e
je
je
K
KPH
(6)
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 417 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
4/12
where RicianHK
K
1+is the LOS component of the
channel and RayleighHK 1
1
+is the fading
component. K is the Rician factor, and P is thepower of each tap. Xij (i-th receiving and j-thtransmitting antenna) are correlated zero-mean, unit
variance, complex Gaussian random variablesconsidered as coefficients of the variable NLOS(Rayleigh) matrix HRayleigh, e
j are the elements ofthe fixed LOS matrixHRician. It is assumed that eachtap consists of a number of individual rays so that
the complex Gaussian assumption is valid.Pin (5)represents the sum of the fixed LOS power and the
variable NLOS power (sum of powers of all taps).Parameters in the 5GHz channel model are as
follows:
Table 1: Channel B parameters for IEEE 802.11n
Model
NameCondition
Rician
factor
K
(dB)
RMS
delay
spread(ns)
No. of
clusters
BLOS and
NLOS/0 15 2
4. MIMO Transceiver for 5 GHz and60 GHzConsidering a MIMO structure containing{1,2,.MT} transmitter antennas and {1,2,MR}receiver antennas , the discrete time channel ismodeled using the equation,
Y=HX + N (7)Where, Y is the (MR x 1) channel output, X is
the (MT x1) transmitted data. H is the (MR x MT)impulse response of the channel which contains theTime of Arrival (ToA) and AoA information and Nis the (MR x 1) Additive White Gaussian Noise(AWGN). The work contributed by us (SavithaManojna et.al[20]), has analysed the MIMOtransceiver performance for 60 GHz. The influenceof TSV and channel B on the multiplexed datastream and equalization of complex channel co-efficient due to time-dispersive channel arepresented in the following two sections.
4.1 Spatial MultiplexingThe multipath fading is considered to be that of the
near-field effect, as the system is designed forindoor environment. With dominance of LOScomponent, spatial multiplexing is considered to
improve the channel capacity. By providingmultiple transmit and multiple receive antennae,
spatial dimension increases and this leads to theincrease in degree-of-freedom [21] [24][25]and[26]. The MIMO concept can be shown in thediagram as follows;
Fig. 1 MIMO Transceiver
The multiplexed data stream X, has both trainingand data. The training symbols train the receiver inidentifying the spatial signatures, which isattenuated due to interference from the adjacentchannel. This is evident from the Fig.1, receiveantenna 1 (RX1), receives data from all transmitantenna TX 1,TX 2....TX Nt. The recovery of the
spatial signature is performed using linear
estimation techniques [27]. MIMO system isconsidered to be a set of parallel SISO systems.The singular value decomposition (SVD) of thechannel matrix H results in independent SISO
channel [28].In any case for MIMO spatial multiplexing the
number of receive antenna must be equal to orgreater than the number of transmit antenna. Totake advantage of the additional throughput
offered, MIMO wireless systems utilize a matrixmathematical approach [29][30]and [31].
Spatial Multiplexing assumes full rank channel
matrix i.e. the spatial correlation arising due to thescatterers, spacing between the antenna elementsand antenna array geometry is low. Hence, MTindependent data symbols are transmitted per
symbol period [10][28]. Considering no temporalinter-symbol interference (ISI), that is made
possible with high directive antenna, the channelmatrix has a dominant LOS. This contributes toimprovement in performance in lower Eb/No range.
With TSV being a two-ray model, one formed byLOS and the other one being the ground reflected
path, the signal propagation is dominated by LOS[8]. In 5 GHz radio, multipath fading contributes tothe major part of the signal propagation. Hence the
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 418 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
5/12
channel impulse response in equation (7) is givenby MRx MT matrix H (t,)
=
),(....),( .
.
.
.
.
.
),(....),(
),(
,1,
,11,1
thth
thth
tH
TRR
T
MMM
M
(8)
The time varying impulse response between jth
(1,2, MT) transmitter antenna and ith (1,2,.MR)
receiver antenna is denoted as hi,j(t,). The MIMOchannel capacity can be calculated for TSV and
channel B by performing SVD on H.The MIMO channel capacity with Mt inputs and
Mr output signals with equal power allocation inthe transmitter as the channel state index (CSI) isunknown at the transmitter is represented as
+= H
T
M HHM
SNRIEC
rdetlog 2 (9)
withRM
I being the identity matrix of size MR x
MT, where SNR is the signal to noise ratio and H isthe channel matrix and H
His the hermitian
transpose.Assuming perfect channel estimation, the
received signal is sent to the different equalizationschemes to estimate the transmitted signal and toreduce the noise. Three equalization schemes i.e.Zero Forcing, Minimum Mean Square Error andMaximum Likelihood are compared and the biterror rate is determined [20].
4.2 Equalization of channel effectsThe following analysis is performed to reduce thechannel impairments and reduce the ISI along withthe noise.
A) Minimization of ISI using Zero Forcingtechnique
The time-dispersive nature of the channel withinthe cluster leads to ISI. The effect of ISI is reducedby using the ZFE. The channel matrix H is afunction of delay spread, angle spread and dopplerspread. The effect of doppler spread in indoorenvironment is too low and hence neglected.With perfect channel estimation, the elements inthe channel matrix can be equalized by multiplyingthe received signal with Wzf= (H*H)
-1 H*. where His the (MR x MT) impulse response of the channelwhich contains the ToA and AoA information and
H* is the conjugate of H.
As a comparison of 60 GHz and 5 GHz withrespect to ZFE, distance of propagation is a matter
of concern. In MMW, oxygen absorption limits thepropagation and with directional antenna the effectof ISI will not be felt [10][32]. When LOScomponent is fully blocked, the received signalpower goes below the receiver sensitivity and is left
undetected. Hence directional antenna withmoderate half power beamwidth is preferred, for
which ZFE is required. In 5 GHz, ISI effect is pre-dominant and is effectively neutralized with one-tap filter.
B) Maximization of SNR using MaximumLikelihood Receiver (ML)
Due to multipath fading and noise, the receivedsignal appears as a cloud of points around the
desired signal. ML performs a comparison of thesepoints with the desired signal and the optimum
signal is determined. The optimum signal is themost likely signal with least error probability [27].
The ML principle is given as
2||||minarg
}....2,1{
xym
Nyyykx
=
(10)
The noise effect in the received signal is takeninto account on computing the most likely path.MMW and 5 GHz radio gain advantage with thisML.
C) Minimization of ISI and noise usingMinimum Mean Square Error Receiver
Compared to ZFE, MMSE minimizes ISI andnoise. With full rank channel, the channelcorrelation HH
*results in identity matrix. But, in
reality, this cannot be achieved. The antenna arraygeometry, scatterer location and antenna
polarization can be considered such that theuncorrelated channels are obtained.
MMSE computes the channel correlation Rh and
effect of noise as *1 1HISNR
RW hMMSE
+= ,
in the received signal in (7).
5. Results and DiscussionThe results obtained in the channel modeling andthe spatial multiplexing of 2x2, 4x4, 8x8 antennaewith 5 GHz and 60 GHz channel models areanalyzed. The bit error rate of this configuration iscomputed and compared for various equalization
techniques i.e. Zero Forcing, Minimum MeanSquare Error and Maximum Likelihood. Also, thecapacity and link budget is presented. At high SNR,
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 419 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
6/12
the channel capacity increases with SNR as min{MT, MR} log SNR (bps/Hz), in contrast to log
SNR for single channels [21] .Thus multipleantenna channels are min { MT, MR } parallelspatial channels, which accounts hence it is thetotal number of degrees of freedom tocommunicate.
5.1 Power delay profile analysis for MMWPower delay profile (PDP) denotes the averagepower that is associated with the given multipathdelay. Using (1), the channel is modeled with therange being 3m and number of clusters consideredto be 3. The LOS desktop parameters provided byNICT, Japan are used as part of two path statisticalmodel [22]. The following parameters are used in
modeling the channel.
Path loss = (4/)2d
nwhere d= 3m, =0.005m,
n=2Therefore, Path loss = 77.56 dB
Table 2: Parameters used in the TSV channelmodel
Channel model Desktop CM 7
Distance between transmitter
and receiver
3m
LOS component path loss 77.54 dB
Average number of clusters 3
Cluster power level - 92 dB
Small Rician effect 2.53
Cluster arrival angle in deg 34.6
Cluster arrival rate (Cluster
per ns)
27
Ray arrival rate (Ray per ns) 1.56
Cluster decay rate (ns) 21.1
Ray decay rate (ns) 8.85
Standard deviation for
lognormal variable for cluster
fading
3.01
Standard deviation for
lognormal variable for Ray
fading
7.69
0 50 100 150
-140
-130
-120
-110
-100
-90
-80
Time of arrival[ns]
Averagepower[d
B]
Fig. 2 PDP of the TSV channel
From Fig 2, the following parameters are calculated
Table 3: Simulation results of the TSV model used
for MMW
Average RMS delay 1.28 [ns]
Maximum RMS delay 0.004 [ns]
Minimum RMS delay 0.015 [ns]
Average Rician factor 38.51 [dB]
Maximum Rician factor 72.99 [dB]
Minimum Rician factor 1.18 [dB]
The PDP in Fig.2 shows the presence of directcomponent having the average power of -80 dB. Asthe number of reflections per ray increases, the
corresponding amplitude of the ray decreasesbecause of both reflection losses and higher free
space losses.
0 2 4 6 8 10 12 14-150
-140
-130
-120
-110
-100
-90
-80
Time of arrival[ns]
Relativepower[dB]
cluster1 w ith AoA 15 degrees
Fig. 3PDP of cluster1 as a function of ToA withAoA 15o
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 420 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
7/12
0 2 4 6 8 10 12 14 16 18-150
-140
-130
-120
-110
-100
-90
-80
Time of arrival[ns]
Relativepow
er[dB]
cluster2 w ith AoA 30 degrees
Fig. 4 PDP of cluster2 as a function of ToA withAoA 30
o
0 5 10 15 20 25-150
-140
-130
-120
-110
-100
-90
-80
Time of arrival[ns]
Relativepower[dB]
cluster3 w ith AoA 60 degrees
Fig. 5 PDP of cluster3 as a function of ToA withAoA 60
o
0 2 4 6 8 10 12-140
-130
-120
-110
-100
-90
-80
Time of arrival[ns]
Relativepo
wer[dB]
cluster4 w ith AoA 120 degrees
Fig. 6 PDP of cluster4 as a function of ToA withAoA 120o
The exponential PDP is evident from Fig. 3,Fig. 4, Fig. 5 and Fig. 6, as the various clusters
have a maximum LOS component and powerdecaying as a function of distance travelled by thesignal. Cluster is formed by scattering and the startand end of cluster depends on the geometry of thescattering obstacle. From the above figures, arrival
rates and decay factors of the cluster and ray aredetermined for computing the impulse response of
the channel as discussed in section 2 for two pathmodel. Also the angular spread required to analyzethe wide-sense stationary uncorrelated scattering(WSSUCS) characteristics of the channel andantenna directivity can be determined. In this, the
ray clustering is evidenced by the peaks in the PDP.The LOS component path gain is also computed[22]. Likewise the delay spread and the rician
factor K can be calculated from ToA and power inthe LOS and NLOS components. Cluster1, 2, 3 and
4 contribute significantly to the multipathcomponents at the receiving end. The angularspread of the clusters is 110
o.
5.2 Bit error rate comparison using ZF,
MMSE and ML equalization for TSV
channelThe various equalization techniques are used toanalyze the bit error rate for different MIMO
configurations using TSV channel. MIMO systemis simulated using the BPSK modulation scheme,the distance of separation between the transmitter
and receiver being 3m and the parameters in Table2 are considered for modeling the channel.
1 2 3 4 5 6 7 8
10-1.9
10-1.8
10-1.7
10-1.6
10-1.5
10-1.4
Average Eb/No, dB
BitEr
rorRate
ZFE
MMSE
ML
Fig. 7 Comparison of BER for 2x2 MIMO systemusing ZFE, MMSE and ML technique.
From Fig.7, it can be inferred that the MLreceiver is optimal in reducing the error probability
and estimates better as it compares the receivedsignal with all the possible transmitted vectors
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 421 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
8/12
while the ZF receiver has the advantage ofcomplexity reduction, but as the multipath
components are large, the noise components are notsuitably taken care of by ZF, hence ZF shows areduced performance. MMSE technique is able toreduce both the interference and noise componentsas compared to ZF even though it has a slight edge
in performance with 2x2 system, as the number ofantennas increase, they perform almost equally.
1 2 3 4 5 6 7 810
-6
10-5
10-4
10-3
10-2
10-1
Average Eb/No,dB
BitErrorRate
ZF
MMSE
ML
Fig. 8 Comparison of BER for 4x4 MIMO systemusing ZFE, MMSE and ML technique.
1 2 3 4 5 6 7 810
-6
10-5
10-4
10-3
10-2
10-1
Average Eb/No,dB
BitErrorRate
ZF
MMSE
ML
Fig. 9 Comparison of BER for 8x8 MIMO system
using ZFE, MMSE and ML technique.
The results obtained for 2x2, 4x4 and 8x8 usingTSV channel model indicate an increase in datarate as (nt.nr) increase. A BER of 10
-5 at SNR = 8dB for 4x4 and 8x8 indicates any further increasein transmit/receive antenna may not provideperformance improvement.
5.3 Bit error rate comparison using ZF,
MMSE and ML equalization for channel BThe following parameters of 2x2, 4x4 and 8x8 areused for channel B
Table 4: Simulation parameters for 5 GHz System
Modulation used BPSK
Number of bits 10
Average Energy per
bit to Noise power
spectral density
(E /No)
1 to 8 dB
Channel model IEEE 802.11n, Channel
Model B
Centre frequency 5 GHz
1 2 3 4 5 6 7 810
-6
10-5
10-4
10-3
10-2
10-1
Average Eb/No,dB
BitErrorR
ate
ZF-SIC
MMSE-SIC
ML-SIC
Fig. 10. BER for a 2x2 spatial multiplexed system
1 2 3 4 5 6 7 810
-6
10-5
10-4
10-3
10-2
10-1
Average Eb/No,dB
BitErrorRate
ZF-SIC
MMSE-SIC
ML-SIC
Fig.11 BER for a 4x4 spatial multiplexed system
1 2 3 4 5 6 7 810-6
10-5
10-4
10-3
10-2
10-1
Average Eb/No,dB
BitErrorRate
ZF-SIC
MMSE-SIC
ML-SIC
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 422 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
9/12
Fig. 12 BER for a 8x8 spatial multiplexed systemThus the BER improves with increase in Eb/No.
The performance of 5GHz using channel B wasstudied for indoor environment. ZF and MMSEachieves a BER of 10-2.5 at 8 dB, while MLachieves BER of 10
-6at 5dB.
5.4 Comparison of BER for 60GHz and 5
GHz using TSV and IEEE 802.11n channel
B
1 2 3 4 5 6 7 810
-6
10-5
10-4
10-3
10-2
10-1
Average Eb/No,dB
B
it
E
rro
r
R
ate
ZF-SIC-5GHz
MMSE-SIC-5GHz
ML-SIC-5GHz
ZF-SIC-60GHz
MMSE-SIC-60GHz
ML-SIC-60GHz
Fig.13 Bit Error Rate of 4x4 system using TSV andRayleigh channel models for indoor environmentwith transceiver separation of 3m.
In Fig 13 and Fig 14 performance of variousequalization techniques assuming perfect channel
estimate is analysed for 4x4 and 8x8 system. TheML equalizer shows performance improvementcompared to ZF and MMSE in both 60 GHz and 5GHz systems.
1 2 3 4 5 6 7 810
-6
10-5
10-4
10-3
10-2
10-1
Average Eb/No,dB
BitErro
rRate
ZF-SIC-5GHz
MMSE-SIC-5GHz
ZF-SIC-60GHz
MMSE-SIC-60GHz
ML-SIC-60GHz
ML-SIC-5GHz
Fig.14 Bit Error Rate of 8x8 system using TSV andRayleigh channel models for indoor environmentwith transceiver separation of 3m.
Comparing Fig.13 and Fig.14 a power gain of 3
dB is achieved in 60 GHz range using ML
technique. The Rician factor K in Table 3 indicatesthe dominance of LOS component and hence
improvement in performance of MMW system forshort range communication.
5.5 Capacity calculation of 5 GHz and 60GHz systemsCapacity calculation:
The capacity of 2x2 MIMO channel is given by
=
+=
r
i
i
T
s
NM
EC
1 0
2 1log (11)
where r is rank of the channel matrix H obtained bySVD, Es/MTN0 is the signal to noise power pertransmit antenna with symbol period T= 1second
and is the Eigen value representing the channelpower gain. Applying SVD on H in (9), the rparallel SISO channels are obtained as in (11)
Table 5: Comparison of capacity for 60 GHz and 5GHz spatial multiplexed systems
Center Frequency Data Rate
60 GHz 2 Gbps
5 GHz 600 Mbps
5.6 Link budget analysis for MMWTo design a communication system for an intendedenvironment, it is necessary to predict the systemperformance before deployment. The essential
parameters that control the performance of eachlink are received signal strength, noise in thereceived signal and channel impairments. For this,link budget is prepared.
Link Margin Figure = Power Received Receiversensitivity (12)
Calculation of Power Received and Receiver
Sensitivity
Data Rate Rb obtained from the simulation is 2Gbps, and the noise bandwidth B is 1 GHz and letEb/No be 8 dB with transmit and receive separation
being 3m.
Carrier to Noise Ratio (CNR) is given as:
=
B
R
N
ECNR b
o
b *log10 10 (13)
= 11 dB
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 423 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
10/12
=
dPathloss
4log20 10 (14)
= 77.54 dB
Power Received=Ptx + Gtx + Grx - Path loss (15)= 10 dBm + 6 dBi + 6 dBi 77.54 (dB)= -45.54 dBm
Calculation of noise power (at T=290 k and
Noise Figure F = 8 dB)
Pnoise = 10log10(kTBF) (16)
= -174 dBm/Hz + 10 log10 (B) + 8 dB= -136 dBm.
Calculation of receiver sensitivity
noiserx PCNRP +=(17)= 11 + (-136) = -105 dBm
Link Margin Figure = Power received - Receiversensitivity
= - 45.54 (-105) = 59.46 dBm
The receiver sensitivity of -105 dBm indicatesthe receiver can process signals with SNR as low as11 dB.
6. ConclusionThe TSV model developed by NICT, Japan isconsidered for 60 GHz. The channel characteristicsof the indoor environment using TSV model is
studied and the PDP is shown. As the scattererslocal to the transmitter contribute to multipath
fading in MMW, the near-field effect is dominantcompared to the far-field effect. The delay spread isreasonably low with average delay spread of 1.289ns and the channel is considered flat fading. Thispaves way for deploying multiple MMW
transceivers in a single room. Performance of the
spatial multiplexed system for 60 GHz and 5 GHzchannel with 2x2, 4x4 and 8x8 were analyzed.With dominance of LOS component in 60 GHz, thebit error rate is reduced and the channel capacity isincreased. ML is found to have reduced BER at lowSNR which suggests that MMW can be the lastmile solution of WLAN operating at 5 GHz. Thedrawback being if the LOS component is blocked,the system will incur loss of data increasing the biterror rate. The future work, is to consider schemesthat will maintain the bit error rate even with loss
of LOS component.
References:[1]C. Liu, E. Skafidas, and R. J. Evans, Capacity
and Data Rate for Millimeter WavelengthSystems in a Short Range Package RadioTransceiver, IEEE Transactions On WirelessCommunications, Vol. 9, No. 3, March 2010,pp -903-906.
[2] S. K. Yong and C.-C. Chong, An overview ofmultigigabit wireless through millimeter wave
technology: Potentials and technicalchallenges, EURASIP J. Wireless Commun.and Networking, Vol. 2007, Article ID 78907,2007.
[3] Johan Karedal, Peter Almers, Anders J
Johansson, Fredrik Tufvesson, and Andreas F.Molisch, A MIMO Channel Model forWireless Personal Area Networks, IEEE
Transactions on Wireless Communications,Vol. 9, No. 1, January 2010, pp 245-255.
[4]Markus Muck, Philippe Bernardin, PatrickLabbe, Xavier Miet, Dirk Pannicke, JensSchonthier, Prototyping of a hybrid5GHz/60GHz OFDM WLAN system in theframework of IST-BroadWay, ISTBROADWAY projectIST-2001-32686, 2002.
[5]Tony K. Mak, Kenneth P. Laberteaux, Raja
Sengupta, and Mustafa Ergen , MultichannelMedium Access Control for Dedicated Short-Range Communications, IEEE Transactions
on Vehicular Technology, Vol. 58, No. 1,January 2009, pp-345-366.
[6] S. Geng, X. Zhao, J. Kivinen, and P.Vainikainen, Millimeter-wave PropagationChannel Characterization for short-rangewireless communications, IEEE Transactionson Vehicular Technology, Vol. 58, January2009, pp. 3- 13.
[7] S. Geng, S. Ranvier, X. Zhao, J. Kivinen, andP. Vainikainen, Multipath PropagationCharacterization of Ultra-Wide Band IndoorRadio Channels, IEEE International
Conference on Ultra- Wideband (ICU05)Zurich 2005, September 5-8, 2005,
[8]Colin Sheldon, Eric Torkildson, Munkyo Seo,C. Patrick Yue, Mark Rodwell, and UpamanyuMadhow, Spatial Multiplexing Over a Line-of-Sight Millimeter-Wave MIMO Link: A a-Channel Hardware Demonstration at 1.2GbpsOver 41m range, Proceedings of EUMA,
October 2008.[9]Colin Sheldon, Munkyo Seo, Eric Torkildson,
Mark Rodwell and Upamanyu Madhow Four-Channel Spatial Multiplexing Over aMillimeter-Wave Line-of-Sight Link, IEEE
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 424 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
11/12
MTT-S International Microwave SymposiumDigest, MTT '09, 2009.
[10] Eric Torkildson, Colin Sheldon, UpamanyuMadhow, Mark Rodwell, Millimeter-WaveSpatial Multiplexing in an IndoorEnvironment,First International Workshop onMulti-Gigabit MM-Wave and Tera-Hz
Wireless Systems (MTWS 09), November2009.
[11] Jan Hansen, A Novel Stochastic Millimeter-Wave Indoor Radio Channel Model, IEEEJournal On Selected Areas InCommunications, Vol. 20, No. 6, August2002, pp 1240-1246.
[12] Oreste Andrisano, Velio Tralli and RobertoVerdone, Millimeter Waves for Short-RangeMultimedia Communication Systems,
Proceedings of The IEEE, Vol. 86, No. 7,July1998, pp-1383-1399.
[13] F. Bohagen, P. Orten, and G. Oien,Construction and capacity analysis of high-rank line-of-sight MIMO channels, IEEEWireless Communications and NetworkingConference, Vol. 1, March 2005, pp. 432437.
[14] Q.H. Spencer, B.D. Jeffs, M.A. Jensen, A.L.Swindlehurst, Modeling the statistical time
and angle of arrival characteristics of anindoor multipath, IEEE Journal on SelectedAreas in Communications, Vol.18, no.3,
March 2000 , pp.347-360.[15] Minyoung Park, Carlos Cordeiro, Eldad
Perahia, and L. Lily Yang, Millimeter-WaveMulti-Gigabit WLAN: Challenges andFeasibility, The Second European Conferenceon Antennas and Propagation Vol. 10, April2008, pp 345 349.
[16] Jingjing Wang, Hao Zhang, Tingting Lv andT. Aaron Gulliver, Capacity of 60 GHzWireless Communication Systems over FadingChannels, Journal of Networks, Academypublisher, Vol. 7, No.1, January 2012.
[17] Hirokazu Sawada, Yozo Shoji and Chang-Soon Choi, Proposal of novel statistic channelmodel for Millimeter Wave WPAN TSV:Shoji-Sawada-Saleh (Triple S)-Valenzuelamodel, Proceedings of Asia-PacificMicrowave Conference 2006.
[18] IEEE P802.11 Wireless LANs, TGn ChannelModels, IEEE 802.11-03/940r4, 2004-05-10.
[19] John S.Seybold, Introduction to RFpropagation,John Wiley and Sons, 2005.
[20] Savitha Manojna.D, Kirthiga.S andJayakumar.M, Spatial Multiplexing forMillimeter Waves Using TSV Model,International Journal of Computer Science
and Engineering Technology Vol.1, Issue 5,ISSN: 2231-0711, June 2011, 244-248.
[21] A. Poon, R. Brodersen, and D. Tse, Degreesof freedom in multiple antenna channels: asignal space approach, IEEE Transactions onInformation Theory, vol. 51, no. 2, February2005, pp. 523536.
[22] Katsuyoshi Sato, Hirokazu Sawada, YozoShoji and Shuzo Kato (NICT, Yokosuka
Japan)Channel Model for Millimeter WaveWPAN, 18
thAnnual IEEE International
Symposium on Personal, Indoor and MobileRadio Communications (PIMRC 07)September 2007.
[23] W. Weichselberger, M. Herdin, H. O zcelik,and E. Bonek, A stochastic MIMO channelmodel with joint correlation of both link
ends, IEEE Transactions on WirelessCommunications., Vol. 5, No. 1, January
2006, pp. 90-100.[24] V. V. Veeravalli, Y. Liang, and A. M.
Sayeed, Correlated MIMO wireless channels:capacity, optimal signaling, and asymptotics,IEEE Transactions on. Information. Theory,Vol. 51, No. 6, June 2005, pp.2058-2072.
[25] K. I. Pedersen, J. B. Andersen, J. P. Kermoal,
and P. E. Mogensen, A stochastic multiple-input multiple-output radio channel model forevaluation of space-time coding algorithms, in
Proc. Vehicular Technology Conf., Boston,MA,. September 2000.
[26] Y. H. Kho and D. P. Taylor, MIMO channelestimation and tracking based on polynomialprediction with application to equalization,IEEE Transactions. Vehicular. Technology.,Vol. 57, No. 3, May 2008, pp. 1585-1595.
[27] M. Biguesh and A. B. Gershman, Training-based MIMO channel estimation: a study ofestimator tradeoffs and optimal trainingsignals, IEEE Trans. Signal Process., Vol. 54,No. 3, March 2006, pp. 884-893.
[28] E. Torkildson, B. Ananthasubramaniam, U.Madhow, and M. Rodwell,Millimeter-waveMIMO: Wireless links at optical speeds,Proceedings. of 44th Allerton Conference onCommunication, Control and Computing,September 2006.
[29] I. Sarris, A. R. Nix, and A. Doufexi, Highperformance WIMAX architecture using
mimo technology in line-of-sight, The SecondEuropean Conference on Antennas andPropagation, November 2007, pp. 15.
[30] C. Hofmann, A. Knopp, D. Ogermann, R.Schwarz, and B. Lankl, Deficiencies ofcommon MIMO channel models with regard
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 425 Issue 11, Volume 11, November 2012
-
7/29/2019 mimo indoor
12/12
to indoor line-of sight channels, IEEE 19thInternational Symposium in Personal, Indoor
and Mobile Radio Communications ( PIMRC08)on, September 2008, pp. 16.
[31] F. Bohagen, P. Orten, and G. Oien, Design ofoptimal high-rank line of-sight MIMOchannels, IEEE Transactions on Wireless
Communications, Vol. 6, No. 4, April 2007,pp. 14201425.
[32]N. Guo, R. C. Qiu, S. S. Mo, K. Takahashi,60-GHz Millimeter-Wave Radio: Principle,Technology and New Results, EURASIPJournal on Wireless Communications andNetworking, 2007.
WSEAS TRANSACTIONS on COMMUNICATIONS S. Kirthiga, M. Jayakumar
E-ISSN: 2224-2864 426 Issue 11, Volume 11, November 2012